Design rainfall estimation in South Africa using Bartlett-Lewis rectangular pulse rainfall models

Relatively few recording raingauges with long reliable records are available in South Africa for the estimation of short duration (less than or equal to 24 h) design rainfalls. Hence, approaches to estimating short duration design rainfalls have been developed for South Africa which are based on daily rainfall data measured by standard, non-recording raingauges at fixed 24 h periods ending at 08:00. These approaches include the use of regional frequency analyses, scaling the moments of the extreme events and stochastic modelling of the rainfall process (Smithers and Schulze, Development and evaluation of techniques for estimating short duration design rainfall in South Africa, WRC Report No. 681/1/00, Water Research Commission, Pretoria, RSA, 2000a pp. 356). In this paper, the methodology of determining the parameters of two versions of Bartlett-Lewis Rectangular Pulse Models (BLRP) and a comparison of the performance of the models at selected stations is presented. The correlation between the parameters of the models is illustrated and utilised in better defining parameters and searching for appropriate model parameters. The scaling characteristics of rainfall are used to estimate the variances for durations <24 h from daily rainfall data, which are then used to determine model parameters. An assessment of the performance of the models with parameters determined using readily available daily rainfall data, and with shorter duration statistics inferred from the daily data, is presented for selected research sites where the short duration rainfall data are deemed to be reliable. The assessments of performance include comparisons between moments and design rainfall depths computed from observed data and from synthetic rainfall series generated by the models. (C) 2002 Elsevier Science B.V. All rights reserved.